© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Natural Language
    Autoencoders
    NEW
    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    2. Google DeepMind · Exploring Gemma 2 with Gemma Scope
    3. Gemma-2-9B
    4. Residual Stream - 16k
    5. 20-GEMMASCOPE-RES-16K
    6. 854
    Prev
    Next
    INDEX
    Explanations

    syntax elements related to code structure

    oai_token-act-pair · gpt-4o-miniTriggered by @bot

    code-related syntax and programming terminology, particularly from technical configuration files and XML/JSON schemas.

    oai_token-act-pair · claude-3-5-haiku-20241022Triggered by @neilrathi

    markers that close or terminate structured constructs in code and data (end-of-blocks, function/command calls, self-closing elements, and array/object closures).

    oai_token-act-pair · gpt-5Triggered by @jyhe0408

    closing braces, brackets, or parentheses in code, especially when followed by semicolons or appearing at the end of code blocks.

    oai_token-act-pair · claude-4-5-sonnetTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GEMMA-2-9B @ 20-gemmascope-res-16k
    Configuration
    google/gemma-scope-9b-pt-res/layer_20/width_16k/average_l0_68
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    16,384
    Data Type
    float32
    Hook Name
    blocks.20.hook_resid_post
    Hook Layer
    20
    Architecture
    jumprelu
    Context Size
    1,024
    Dataset
    monology/pile-uncopyrighted
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     EconPapers
    -0.73
    __*/
    -0.66
     snippetHide
    -0.64
    RenderAtEndOf
    -0.64
    KommentareTeilen
    -0.64
    Чыгана
    -0.63
    
    -0.62
    <unused28>
    -0.62
    [@BOS@]
    -0.62
    <pad>
    -0.62
    POSITIVE LOGITS
    vfill
    0.32
     acompaña
    0.30
    สำ
    0.29
    šanai
    0.28
    Secondly
    0.28
     emperador
    0.28
     départ
    0.28
    Recv
    0.27
    appui
    0.27
     présence
    0.27
    Activations Density 0.378%

    No Known Activations